2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/hardirq.h>
44 #include <linux/kernel.h>
45 #include <linux/list.h>
46 #include <linux/list_nulls.h>
47 #include <linux/timer.h>
48 #include <linux/cache.h>
49 #include <linux/lockdep.h>
50 #include <linux/netdevice.h>
51 #include <linux/skbuff.h> /* struct sk_buff */
53 #include <linux/security.h>
54 #include <linux/slab.h>
55 #include <linux/uaccess.h>
57 #include <linux/filter.h>
58 #include <linux/rculist_nulls.h>
59 #include <linux/poll.h>
61 #include <linux/atomic.h>
63 #include <net/checksum.h>
66 * This structure really needs to be cleaned up.
67 * Most of it is for TCP, and not used by any of
68 * the other protocols.
71 /* Define this to get the SOCK_DBG debugging facility. */
72 #define SOCK_DEBUGGING
74 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
75 printk(KERN_DEBUG msg); } while (0)
77 /* Validate arguments and do nothing */
78 static inline __printf(2, 3)
79 void SOCK_DEBUG(struct sock
*sk
, const char *msg
, ...)
84 /* This is the per-socket lock. The spinlock provides a synchronization
85 * between user contexts and software interrupt processing, whereas the
86 * mini-semaphore synchronizes multiple users amongst themselves.
93 * We express the mutex-alike socket_lock semantics
94 * to the lock validator by explicitly managing
95 * the slock as a lock variant (in addition to
98 #ifdef CONFIG_DEBUG_LOCK_ALLOC
99 struct lockdep_map dep_map
;
108 * struct sock_common - minimal network layer representation of sockets
109 * @skc_daddr: Foreign IPv4 addr
110 * @skc_rcv_saddr: Bound local IPv4 addr
111 * @skc_hash: hash value used with various protocol lookup tables
112 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
113 * @skc_family: network address family
114 * @skc_state: Connection state
115 * @skc_reuse: %SO_REUSEADDR setting
116 * @skc_bound_dev_if: bound device index if != 0
117 * @skc_bind_node: bind hash linkage for various protocol lookup tables
118 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
119 * @skc_prot: protocol handlers inside a network family
120 * @skc_net: reference to the network namespace of this socket
121 * @skc_node: main hash linkage for various protocol lookup tables
122 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
123 * @skc_tx_queue_mapping: tx queue number for this connection
124 * @skc_refcnt: reference count
126 * This is the minimal network layer representation of sockets, the header
127 * for struct sock and struct inet_timewait_sock.
130 /* skc_daddr and skc_rcv_saddr must be grouped :
131 * cf INET_MATCH() and INET_TW_MATCH()
134 __be32 skc_rcv_saddr
;
137 unsigned int skc_hash
;
138 __u16 skc_u16hashes
[2];
140 unsigned short skc_family
;
141 volatile unsigned char skc_state
;
142 unsigned char skc_reuse
;
143 int skc_bound_dev_if
;
145 struct hlist_node skc_bind_node
;
146 struct hlist_nulls_node skc_portaddr_node
;
148 struct proto
*skc_prot
;
153 * fields between dontcopy_begin/dontcopy_end
154 * are not copied in sock_copy()
157 int skc_dontcopy_begin
[0];
160 struct hlist_node skc_node
;
161 struct hlist_nulls_node skc_nulls_node
;
163 int skc_tx_queue_mapping
;
166 int skc_dontcopy_end
[0];
171 * struct sock - network layer representation of sockets
172 * @__sk_common: shared layout with inet_timewait_sock
173 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
174 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
175 * @sk_lock: synchronizer
176 * @sk_rcvbuf: size of receive buffer in bytes
177 * @sk_wq: sock wait queue and async head
178 * @sk_dst_cache: destination cache
179 * @sk_dst_lock: destination cache lock
180 * @sk_policy: flow policy
181 * @sk_receive_queue: incoming packets
182 * @sk_wmem_alloc: transmit queue bytes committed
183 * @sk_write_queue: Packet sending queue
184 * @sk_async_wait_queue: DMA copied packets
185 * @sk_omem_alloc: "o" is "option" or "other"
186 * @sk_wmem_queued: persistent queue size
187 * @sk_forward_alloc: space allocated forward
188 * @sk_allocation: allocation mode
189 * @sk_sndbuf: size of send buffer in bytes
190 * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
191 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
192 * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
193 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
194 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
195 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
196 * @sk_gso_max_size: Maximum GSO segment size to build
197 * @sk_lingertime: %SO_LINGER l_linger setting
198 * @sk_backlog: always used with the per-socket spinlock held
199 * @sk_callback_lock: used with the callbacks in the end of this struct
200 * @sk_error_queue: rarely used
201 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
202 * IPV6_ADDRFORM for instance)
203 * @sk_err: last error
204 * @sk_err_soft: errors that don't cause failure but are the cause of a
205 * persistent failure not just 'timed out'
206 * @sk_drops: raw/udp drops counter
207 * @sk_ack_backlog: current listen backlog
208 * @sk_max_ack_backlog: listen backlog set in listen()
209 * @sk_priority: %SO_PRIORITY setting
210 * @sk_type: socket type (%SOCK_STREAM, etc)
211 * @sk_protocol: which protocol this socket belongs in this network family
212 * @sk_peer_pid: &struct pid for this socket's peer
213 * @sk_peer_cred: %SO_PEERCRED setting
214 * @sk_rcvlowat: %SO_RCVLOWAT setting
215 * @sk_rcvtimeo: %SO_RCVTIMEO setting
216 * @sk_sndtimeo: %SO_SNDTIMEO setting
217 * @sk_rxhash: flow hash received from netif layer
218 * @sk_filter: socket filtering instructions
219 * @sk_protinfo: private area, net family specific, when not using slab
220 * @sk_timer: sock cleanup timer
221 * @sk_stamp: time stamp of last packet received
222 * @sk_socket: Identd and reporting IO signals
223 * @sk_user_data: RPC layer private data
224 * @sk_sndmsg_page: cached page for sendmsg
225 * @sk_sndmsg_off: cached offset for sendmsg
226 * @sk_send_head: front of stuff to transmit
227 * @sk_security: used by security modules
228 * @sk_mark: generic packet mark
229 * @sk_classid: this socket's cgroup classid
230 * @sk_write_pending: a write to stream socket waits to start
231 * @sk_state_change: callback to indicate change in the state of the sock
232 * @sk_data_ready: callback to indicate there is data to be processed
233 * @sk_write_space: callback to indicate there is bf sending space available
234 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
235 * @sk_backlog_rcv: callback to process the backlog
236 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
240 * Now struct inet_timewait_sock also uses sock_common, so please just
241 * don't add nothing before this first member (__sk_common) --acme
243 struct sock_common __sk_common
;
244 #define sk_node __sk_common.skc_node
245 #define sk_nulls_node __sk_common.skc_nulls_node
246 #define sk_refcnt __sk_common.skc_refcnt
247 #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
249 #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
250 #define sk_dontcopy_end __sk_common.skc_dontcopy_end
251 #define sk_hash __sk_common.skc_hash
252 #define sk_family __sk_common.skc_family
253 #define sk_state __sk_common.skc_state
254 #define sk_reuse __sk_common.skc_reuse
255 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
256 #define sk_bind_node __sk_common.skc_bind_node
257 #define sk_prot __sk_common.skc_prot
258 #define sk_net __sk_common.skc_net
259 socket_lock_t sk_lock
;
260 struct sk_buff_head sk_receive_queue
;
262 * The backlog queue is special, it is always used with
263 * the per-socket spinlock held and requires low latency
264 * access. Therefore we special case it's implementation.
265 * Note : rmem_alloc is in this structure to fill a hole
266 * on 64bit arches, not because its logically part of
272 struct sk_buff
*head
;
273 struct sk_buff
*tail
;
275 #define sk_rmem_alloc sk_backlog.rmem_alloc
276 int sk_forward_alloc
;
283 struct sk_filter __rcu
*sk_filter
;
284 struct socket_wq __rcu
*sk_wq
;
286 #ifdef CONFIG_NET_DMA
287 struct sk_buff_head sk_async_wait_queue
;
291 struct xfrm_policy
*sk_policy
[2];
293 unsigned long sk_flags
;
294 struct dst_entry
*sk_dst_cache
;
295 spinlock_t sk_dst_lock
;
296 atomic_t sk_wmem_alloc
;
297 atomic_t sk_omem_alloc
;
299 struct sk_buff_head sk_write_queue
;
300 kmemcheck_bitfield_begin(flags
);
301 unsigned int sk_shutdown
: 2,
306 kmemcheck_bitfield_end(flags
);
309 netdev_features_t sk_route_caps
;
310 netdev_features_t sk_route_nocaps
;
312 unsigned int sk_gso_max_size
;
314 unsigned long sk_lingertime
;
315 struct sk_buff_head sk_error_queue
;
316 struct proto
*sk_prot_creator
;
317 rwlock_t sk_callback_lock
;
320 unsigned short sk_ack_backlog
;
321 unsigned short sk_max_ack_backlog
;
323 struct pid
*sk_peer_pid
;
324 const struct cred
*sk_peer_cred
;
328 struct timer_list sk_timer
;
330 struct socket
*sk_socket
;
332 struct page
*sk_sndmsg_page
;
333 struct sk_buff
*sk_send_head
;
335 int sk_write_pending
;
336 #ifdef CONFIG_SECURITY
341 void (*sk_state_change
)(struct sock
*sk
);
342 void (*sk_data_ready
)(struct sock
*sk
, int bytes
);
343 void (*sk_write_space
)(struct sock
*sk
);
344 void (*sk_error_report
)(struct sock
*sk
);
345 int (*sk_backlog_rcv
)(struct sock
*sk
,
346 struct sk_buff
*skb
);
347 void (*sk_destruct
)(struct sock
*sk
);
351 * Hashed lists helper routines
353 static inline struct sock
*sk_entry(const struct hlist_node
*node
)
355 return hlist_entry(node
, struct sock
, sk_node
);
358 static inline struct sock
*__sk_head(const struct hlist_head
*head
)
360 return hlist_entry(head
->first
, struct sock
, sk_node
);
363 static inline struct sock
*sk_head(const struct hlist_head
*head
)
365 return hlist_empty(head
) ? NULL
: __sk_head(head
);
368 static inline struct sock
*__sk_nulls_head(const struct hlist_nulls_head
*head
)
370 return hlist_nulls_entry(head
->first
, struct sock
, sk_nulls_node
);
373 static inline struct sock
*sk_nulls_head(const struct hlist_nulls_head
*head
)
375 return hlist_nulls_empty(head
) ? NULL
: __sk_nulls_head(head
);
378 static inline struct sock
*sk_next(const struct sock
*sk
)
380 return sk
->sk_node
.next
?
381 hlist_entry(sk
->sk_node
.next
, struct sock
, sk_node
) : NULL
;
384 static inline struct sock
*sk_nulls_next(const struct sock
*sk
)
386 return (!is_a_nulls(sk
->sk_nulls_node
.next
)) ?
387 hlist_nulls_entry(sk
->sk_nulls_node
.next
,
388 struct sock
, sk_nulls_node
) :
392 static inline int sk_unhashed(const struct sock
*sk
)
394 return hlist_unhashed(&sk
->sk_node
);
397 static inline int sk_hashed(const struct sock
*sk
)
399 return !sk_unhashed(sk
);
402 static __inline__
void sk_node_init(struct hlist_node
*node
)
407 static __inline__
void sk_nulls_node_init(struct hlist_nulls_node
*node
)
412 static __inline__
void __sk_del_node(struct sock
*sk
)
414 __hlist_del(&sk
->sk_node
);
417 /* NB: equivalent to hlist_del_init_rcu */
418 static __inline__
int __sk_del_node_init(struct sock
*sk
)
422 sk_node_init(&sk
->sk_node
);
428 /* Grab socket reference count. This operation is valid only
429 when sk is ALREADY grabbed f.e. it is found in hash table
430 or a list and the lookup is made under lock preventing hash table
434 static inline void sock_hold(struct sock
*sk
)
436 atomic_inc(&sk
->sk_refcnt
);
439 /* Ungrab socket in the context, which assumes that socket refcnt
440 cannot hit zero, f.e. it is true in context of any socketcall.
442 static inline void __sock_put(struct sock
*sk
)
444 atomic_dec(&sk
->sk_refcnt
);
447 static __inline__
int sk_del_node_init(struct sock
*sk
)
449 int rc
= __sk_del_node_init(sk
);
452 /* paranoid for a while -acme */
453 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
458 #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
460 static __inline__
int __sk_nulls_del_node_init_rcu(struct sock
*sk
)
463 hlist_nulls_del_init_rcu(&sk
->sk_nulls_node
);
469 static __inline__
int sk_nulls_del_node_init_rcu(struct sock
*sk
)
471 int rc
= __sk_nulls_del_node_init_rcu(sk
);
474 /* paranoid for a while -acme */
475 WARN_ON(atomic_read(&sk
->sk_refcnt
) == 1);
481 static __inline__
void __sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
483 hlist_add_head(&sk
->sk_node
, list
);
486 static __inline__
void sk_add_node(struct sock
*sk
, struct hlist_head
*list
)
489 __sk_add_node(sk
, list
);
492 static __inline__
void sk_add_node_rcu(struct sock
*sk
, struct hlist_head
*list
)
495 hlist_add_head_rcu(&sk
->sk_node
, list
);
498 static __inline__
void __sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
500 hlist_nulls_add_head_rcu(&sk
->sk_nulls_node
, list
);
503 static __inline__
void sk_nulls_add_node_rcu(struct sock
*sk
, struct hlist_nulls_head
*list
)
506 __sk_nulls_add_node_rcu(sk
, list
);
509 static __inline__
void __sk_del_bind_node(struct sock
*sk
)
511 __hlist_del(&sk
->sk_bind_node
);
514 static __inline__
void sk_add_bind_node(struct sock
*sk
,
515 struct hlist_head
*list
)
517 hlist_add_head(&sk
->sk_bind_node
, list
);
520 #define sk_for_each(__sk, node, list) \
521 hlist_for_each_entry(__sk, node, list, sk_node)
522 #define sk_for_each_rcu(__sk, node, list) \
523 hlist_for_each_entry_rcu(__sk, node, list, sk_node)
524 #define sk_nulls_for_each(__sk, node, list) \
525 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
526 #define sk_nulls_for_each_rcu(__sk, node, list) \
527 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
528 #define sk_for_each_from(__sk, node) \
529 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
530 hlist_for_each_entry_from(__sk, node, sk_node)
531 #define sk_nulls_for_each_from(__sk, node) \
532 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
533 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
534 #define sk_for_each_safe(__sk, node, tmp, list) \
535 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
536 #define sk_for_each_bound(__sk, node, list) \
537 hlist_for_each_entry(__sk, node, list, sk_bind_node)
550 SOCK_USE_WRITE_QUEUE
, /* whether to call sk->sk_write_space in sock_wfree */
551 SOCK_DBG
, /* %SO_DEBUG setting */
552 SOCK_RCVTSTAMP
, /* %SO_TIMESTAMP setting */
553 SOCK_RCVTSTAMPNS
, /* %SO_TIMESTAMPNS setting */
554 SOCK_LOCALROUTE
, /* route locally only, %SO_DONTROUTE setting */
555 SOCK_QUEUE_SHRUNK
, /* write queue has been shrunk recently */
556 SOCK_TIMESTAMPING_TX_HARDWARE
, /* %SOF_TIMESTAMPING_TX_HARDWARE */
557 SOCK_TIMESTAMPING_TX_SOFTWARE
, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
558 SOCK_TIMESTAMPING_RX_HARDWARE
, /* %SOF_TIMESTAMPING_RX_HARDWARE */
559 SOCK_TIMESTAMPING_RX_SOFTWARE
, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
560 SOCK_TIMESTAMPING_SOFTWARE
, /* %SOF_TIMESTAMPING_SOFTWARE */
561 SOCK_TIMESTAMPING_RAW_HARDWARE
, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
562 SOCK_TIMESTAMPING_SYS_HARDWARE
, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
563 SOCK_FASYNC
, /* fasync() active */
565 SOCK_ZEROCOPY
, /* buffers from userspace */
566 SOCK_WIFI_STATUS
, /* push wifi status to userspace */
569 static inline void sock_copy_flags(struct sock
*nsk
, struct sock
*osk
)
571 nsk
->sk_flags
= osk
->sk_flags
;
574 static inline void sock_set_flag(struct sock
*sk
, enum sock_flags flag
)
576 __set_bit(flag
, &sk
->sk_flags
);
579 static inline void sock_reset_flag(struct sock
*sk
, enum sock_flags flag
)
581 __clear_bit(flag
, &sk
->sk_flags
);
584 static inline int sock_flag(struct sock
*sk
, enum sock_flags flag
)
586 return test_bit(flag
, &sk
->sk_flags
);
589 static inline void sk_acceptq_removed(struct sock
*sk
)
591 sk
->sk_ack_backlog
--;
594 static inline void sk_acceptq_added(struct sock
*sk
)
596 sk
->sk_ack_backlog
++;
599 static inline int sk_acceptq_is_full(struct sock
*sk
)
601 return sk
->sk_ack_backlog
> sk
->sk_max_ack_backlog
;
605 * Compute minimal free write space needed to queue new packets.
607 static inline int sk_stream_min_wspace(struct sock
*sk
)
609 return sk
->sk_wmem_queued
>> 1;
612 static inline int sk_stream_wspace(struct sock
*sk
)
614 return sk
->sk_sndbuf
- sk
->sk_wmem_queued
;
617 extern void sk_stream_write_space(struct sock
*sk
);
619 static inline int sk_stream_memory_free(struct sock
*sk
)
621 return sk
->sk_wmem_queued
< sk
->sk_sndbuf
;
624 /* OOB backlog add */
625 static inline void __sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
627 /* dont let skb dst not refcounted, we are going to leave rcu lock */
630 if (!sk
->sk_backlog
.tail
)
631 sk
->sk_backlog
.head
= skb
;
633 sk
->sk_backlog
.tail
->next
= skb
;
635 sk
->sk_backlog
.tail
= skb
;
640 * Take into account size of receive queue and backlog queue
642 static inline bool sk_rcvqueues_full(const struct sock
*sk
, const struct sk_buff
*skb
)
644 unsigned int qsize
= sk
->sk_backlog
.len
+ atomic_read(&sk
->sk_rmem_alloc
);
646 return qsize
+ skb
->truesize
> sk
->sk_rcvbuf
;
649 /* The per-socket spinlock must be held here. */
650 static inline __must_check
int sk_add_backlog(struct sock
*sk
, struct sk_buff
*skb
)
652 if (sk_rcvqueues_full(sk
, skb
))
655 __sk_add_backlog(sk
, skb
);
656 sk
->sk_backlog
.len
+= skb
->truesize
;
660 static inline int sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
662 return sk
->sk_backlog_rcv(sk
, skb
);
665 static inline void sock_rps_record_flow(const struct sock
*sk
)
668 struct rps_sock_flow_table
*sock_flow_table
;
671 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
672 rps_record_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
677 static inline void sock_rps_reset_flow(const struct sock
*sk
)
680 struct rps_sock_flow_table
*sock_flow_table
;
683 sock_flow_table
= rcu_dereference(rps_sock_flow_table
);
684 rps_reset_sock_flow(sock_flow_table
, sk
->sk_rxhash
);
689 static inline void sock_rps_save_rxhash(struct sock
*sk
,
690 const struct sk_buff
*skb
)
693 if (unlikely(sk
->sk_rxhash
!= skb
->rxhash
)) {
694 sock_rps_reset_flow(sk
);
695 sk
->sk_rxhash
= skb
->rxhash
;
700 static inline void sock_rps_reset_rxhash(struct sock
*sk
)
703 sock_rps_reset_flow(sk
);
708 #define sk_wait_event(__sk, __timeo, __condition) \
710 release_sock(__sk); \
711 __rc = __condition; \
713 *(__timeo) = schedule_timeout(*(__timeo)); \
716 __rc = __condition; \
720 extern int sk_stream_wait_connect(struct sock
*sk
, long *timeo_p
);
721 extern int sk_stream_wait_memory(struct sock
*sk
, long *timeo_p
);
722 extern void sk_stream_wait_close(struct sock
*sk
, long timeo_p
);
723 extern int sk_stream_error(struct sock
*sk
, int flags
, int err
);
724 extern void sk_stream_kill_queues(struct sock
*sk
);
726 extern int sk_wait_data(struct sock
*sk
, long *timeo
);
728 struct request_sock_ops
;
729 struct timewait_sock_ops
;
730 struct inet_hashinfo
;
734 /* Networking protocol blocks we attach to sockets.
735 * socket layer -> transport layer interface
736 * transport -> network interface is defined by struct inet_proto
739 void (*close
)(struct sock
*sk
,
741 int (*connect
)(struct sock
*sk
,
742 struct sockaddr
*uaddr
,
744 int (*disconnect
)(struct sock
*sk
, int flags
);
746 struct sock
* (*accept
) (struct sock
*sk
, int flags
, int *err
);
748 int (*ioctl
)(struct sock
*sk
, int cmd
,
750 int (*init
)(struct sock
*sk
);
751 void (*destroy
)(struct sock
*sk
);
752 void (*shutdown
)(struct sock
*sk
, int how
);
753 int (*setsockopt
)(struct sock
*sk
, int level
,
754 int optname
, char __user
*optval
,
755 unsigned int optlen
);
756 int (*getsockopt
)(struct sock
*sk
, int level
,
757 int optname
, char __user
*optval
,
760 int (*compat_setsockopt
)(struct sock
*sk
,
762 int optname
, char __user
*optval
,
763 unsigned int optlen
);
764 int (*compat_getsockopt
)(struct sock
*sk
,
766 int optname
, char __user
*optval
,
768 int (*compat_ioctl
)(struct sock
*sk
,
769 unsigned int cmd
, unsigned long arg
);
771 int (*sendmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
772 struct msghdr
*msg
, size_t len
);
773 int (*recvmsg
)(struct kiocb
*iocb
, struct sock
*sk
,
775 size_t len
, int noblock
, int flags
,
777 int (*sendpage
)(struct sock
*sk
, struct page
*page
,
778 int offset
, size_t size
, int flags
);
779 int (*bind
)(struct sock
*sk
,
780 struct sockaddr
*uaddr
, int addr_len
);
782 int (*backlog_rcv
) (struct sock
*sk
,
783 struct sk_buff
*skb
);
785 /* Keeping track of sk's, looking them up, and port selection methods. */
786 void (*hash
)(struct sock
*sk
);
787 void (*unhash
)(struct sock
*sk
);
788 void (*rehash
)(struct sock
*sk
);
789 int (*get_port
)(struct sock
*sk
, unsigned short snum
);
790 void (*clear_sk
)(struct sock
*sk
, int size
);
792 /* Keeping track of sockets in use */
793 #ifdef CONFIG_PROC_FS
794 unsigned int inuse_idx
;
797 /* Memory pressure */
798 void (*enter_memory_pressure
)(struct sock
*sk
);
799 atomic_long_t
*memory_allocated
; /* Current allocated memory. */
800 struct percpu_counter
*sockets_allocated
; /* Current number of sockets. */
802 * Pressure flag: try to collapse.
803 * Technical note: it is used by multiple contexts non atomically.
804 * All the __sk_mem_schedule() is of this nature: accounting
805 * is strict, actions are advisory and have some latency.
807 int *memory_pressure
;
814 struct kmem_cache
*slab
;
815 unsigned int obj_size
;
818 struct percpu_counter
*orphan_count
;
820 struct request_sock_ops
*rsk_prot
;
821 struct timewait_sock_ops
*twsk_prot
;
824 struct inet_hashinfo
*hashinfo
;
825 struct udp_table
*udp_table
;
826 struct raw_hashinfo
*raw_hash
;
829 struct module
*owner
;
833 struct list_head node
;
834 #ifdef SOCK_REFCNT_DEBUG
839 extern int proto_register(struct proto
*prot
, int alloc_slab
);
840 extern void proto_unregister(struct proto
*prot
);
842 #ifdef SOCK_REFCNT_DEBUG
843 static inline void sk_refcnt_debug_inc(struct sock
*sk
)
845 atomic_inc(&sk
->sk_prot
->socks
);
848 static inline void sk_refcnt_debug_dec(struct sock
*sk
)
850 atomic_dec(&sk
->sk_prot
->socks
);
851 printk(KERN_DEBUG
"%s socket %p released, %d are still alive\n",
852 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_prot
->socks
));
855 static inline void sk_refcnt_debug_release(const struct sock
*sk
)
857 if (atomic_read(&sk
->sk_refcnt
) != 1)
858 printk(KERN_DEBUG
"Destruction of the %s socket %p delayed, refcnt=%d\n",
859 sk
->sk_prot
->name
, sk
, atomic_read(&sk
->sk_refcnt
));
861 #else /* SOCK_REFCNT_DEBUG */
862 #define sk_refcnt_debug_inc(sk) do { } while (0)
863 #define sk_refcnt_debug_dec(sk) do { } while (0)
864 #define sk_refcnt_debug_release(sk) do { } while (0)
865 #endif /* SOCK_REFCNT_DEBUG */
868 #ifdef CONFIG_PROC_FS
869 /* Called with local bh disabled */
870 extern void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int inc
);
871 extern int sock_prot_inuse_get(struct net
*net
, struct proto
*proto
);
873 static void inline sock_prot_inuse_add(struct net
*net
, struct proto
*prot
,
880 /* With per-bucket locks this operation is not-atomic, so that
881 * this version is not worse.
883 static inline void __sk_prot_rehash(struct sock
*sk
)
885 sk
->sk_prot
->unhash(sk
);
886 sk
->sk_prot
->hash(sk
);
889 void sk_prot_clear_portaddr_nulls(struct sock
*sk
, int size
);
891 /* About 10 seconds */
892 #define SOCK_DESTROY_TIME (10*HZ)
894 /* Sockets 0-1023 can't be bound to unless you are superuser */
895 #define PROT_SOCK 1024
897 #define SHUTDOWN_MASK 3
898 #define RCV_SHUTDOWN 1
899 #define SEND_SHUTDOWN 2
901 #define SOCK_SNDBUF_LOCK 1
902 #define SOCK_RCVBUF_LOCK 2
903 #define SOCK_BINDADDR_LOCK 4
904 #define SOCK_BINDPORT_LOCK 8
906 /* sock_iocb: used to kick off async processing of socket ios */
908 struct list_head list
;
914 struct scm_cookie
*scm
;
915 struct msghdr
*msg
, async_msg
;
919 static inline struct sock_iocb
*kiocb_to_siocb(struct kiocb
*iocb
)
921 return (struct sock_iocb
*)iocb
->private;
924 static inline struct kiocb
*siocb_to_kiocb(struct sock_iocb
*si
)
929 struct socket_alloc
{
930 struct socket socket
;
931 struct inode vfs_inode
;
934 static inline struct socket
*SOCKET_I(struct inode
*inode
)
936 return &container_of(inode
, struct socket_alloc
, vfs_inode
)->socket
;
939 static inline struct inode
*SOCK_INODE(struct socket
*socket
)
941 return &container_of(socket
, struct socket_alloc
, socket
)->vfs_inode
;
945 * Functions for memory accounting
947 extern int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
);
948 extern void __sk_mem_reclaim(struct sock
*sk
);
950 #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
951 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
952 #define SK_MEM_SEND 0
953 #define SK_MEM_RECV 1
955 static inline int sk_mem_pages(int amt
)
957 return (amt
+ SK_MEM_QUANTUM
- 1) >> SK_MEM_QUANTUM_SHIFT
;
960 static inline int sk_has_account(struct sock
*sk
)
962 /* return true if protocol supports memory accounting */
963 return !!sk
->sk_prot
->memory_allocated
;
966 static inline int sk_wmem_schedule(struct sock
*sk
, int size
)
968 if (!sk_has_account(sk
))
970 return size
<= sk
->sk_forward_alloc
||
971 __sk_mem_schedule(sk
, size
, SK_MEM_SEND
);
974 static inline int sk_rmem_schedule(struct sock
*sk
, int size
)
976 if (!sk_has_account(sk
))
978 return size
<= sk
->sk_forward_alloc
||
979 __sk_mem_schedule(sk
, size
, SK_MEM_RECV
);
982 static inline void sk_mem_reclaim(struct sock
*sk
)
984 if (!sk_has_account(sk
))
986 if (sk
->sk_forward_alloc
>= SK_MEM_QUANTUM
)
987 __sk_mem_reclaim(sk
);
990 static inline void sk_mem_reclaim_partial(struct sock
*sk
)
992 if (!sk_has_account(sk
))
994 if (sk
->sk_forward_alloc
> SK_MEM_QUANTUM
)
995 __sk_mem_reclaim(sk
);
998 static inline void sk_mem_charge(struct sock
*sk
, int size
)
1000 if (!sk_has_account(sk
))
1002 sk
->sk_forward_alloc
-= size
;
1005 static inline void sk_mem_uncharge(struct sock
*sk
, int size
)
1007 if (!sk_has_account(sk
))
1009 sk
->sk_forward_alloc
+= size
;
1012 static inline void sk_wmem_free_skb(struct sock
*sk
, struct sk_buff
*skb
)
1014 sock_set_flag(sk
, SOCK_QUEUE_SHRUNK
);
1015 sk
->sk_wmem_queued
-= skb
->truesize
;
1016 sk_mem_uncharge(sk
, skb
->truesize
);
1020 /* Used by processes to "lock" a socket state, so that
1021 * interrupts and bottom half handlers won't change it
1022 * from under us. It essentially blocks any incoming
1023 * packets, so that we won't get any new data or any
1024 * packets that change the state of the socket.
1026 * While locked, BH processing will add new packets to
1027 * the backlog queue. This queue is processed by the
1028 * owner of the socket lock right before it is released.
1030 * Since ~2.3.5 it is also exclusive sleep lock serializing
1031 * accesses from user process context.
1033 #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
1036 * Macro so as to not evaluate some arguments when
1037 * lockdep is not enabled.
1039 * Mark both the sk_lock and the sk_lock.slock as a
1040 * per-address-family lock class.
1042 #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
1044 sk->sk_lock.owned = 0; \
1045 init_waitqueue_head(&sk->sk_lock.wq); \
1046 spin_lock_init(&(sk)->sk_lock.slock); \
1047 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1048 sizeof((sk)->sk_lock)); \
1049 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
1051 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1054 extern void lock_sock_nested(struct sock
*sk
, int subclass
);
1056 static inline void lock_sock(struct sock
*sk
)
1058 lock_sock_nested(sk
, 0);
1061 extern void release_sock(struct sock
*sk
);
1063 /* BH context may only use the following locking interface. */
1064 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
1065 #define bh_lock_sock_nested(__sk) \
1066 spin_lock_nested(&((__sk)->sk_lock.slock), \
1067 SINGLE_DEPTH_NESTING)
1068 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1070 extern bool lock_sock_fast(struct sock
*sk
);
1072 * unlock_sock_fast - complement of lock_sock_fast
1076 * fast unlock socket for user context.
1077 * If slow mode is on, we call regular release_sock()
1079 static inline void unlock_sock_fast(struct sock
*sk
, bool slow
)
1084 spin_unlock_bh(&sk
->sk_lock
.slock
);
1088 extern struct sock
*sk_alloc(struct net
*net
, int family
,
1090 struct proto
*prot
);
1091 extern void sk_free(struct sock
*sk
);
1092 extern void sk_release_kernel(struct sock
*sk
);
1093 extern struct sock
*sk_clone_lock(const struct sock
*sk
,
1094 const gfp_t priority
);
1096 extern struct sk_buff
*sock_wmalloc(struct sock
*sk
,
1097 unsigned long size
, int force
,
1099 extern struct sk_buff
*sock_rmalloc(struct sock
*sk
,
1100 unsigned long size
, int force
,
1102 extern void sock_wfree(struct sk_buff
*skb
);
1103 extern void sock_rfree(struct sk_buff
*skb
);
1105 extern int sock_setsockopt(struct socket
*sock
, int level
,
1106 int op
, char __user
*optval
,
1107 unsigned int optlen
);
1109 extern int sock_getsockopt(struct socket
*sock
, int level
,
1110 int op
, char __user
*optval
,
1111 int __user
*optlen
);
1112 extern struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
,
1116 extern struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
,
1117 unsigned long header_len
,
1118 unsigned long data_len
,
1121 extern void *sock_kmalloc(struct sock
*sk
, int size
,
1123 extern void sock_kfree_s(struct sock
*sk
, void *mem
, int size
);
1124 extern void sk_send_sigurg(struct sock
*sk
);
1126 #ifdef CONFIG_CGROUPS
1127 extern void sock_update_classid(struct sock
*sk
);
1129 static inline void sock_update_classid(struct sock
*sk
)
1135 * Functions to fill in entries in struct proto_ops when a protocol
1136 * does not implement a particular function.
1138 extern int sock_no_bind(struct socket
*,
1139 struct sockaddr
*, int);
1140 extern int sock_no_connect(struct socket
*,
1141 struct sockaddr
*, int, int);
1142 extern int sock_no_socketpair(struct socket
*,
1144 extern int sock_no_accept(struct socket
*,
1145 struct socket
*, int);
1146 extern int sock_no_getname(struct socket
*,
1147 struct sockaddr
*, int *, int);
1148 extern unsigned int sock_no_poll(struct file
*, struct socket
*,
1149 struct poll_table_struct
*);
1150 extern int sock_no_ioctl(struct socket
*, unsigned int,
1152 extern int sock_no_listen(struct socket
*, int);
1153 extern int sock_no_shutdown(struct socket
*, int);
1154 extern int sock_no_getsockopt(struct socket
*, int , int,
1155 char __user
*, int __user
*);
1156 extern int sock_no_setsockopt(struct socket
*, int, int,
1157 char __user
*, unsigned int);
1158 extern int sock_no_sendmsg(struct kiocb
*, struct socket
*,
1159 struct msghdr
*, size_t);
1160 extern int sock_no_recvmsg(struct kiocb
*, struct socket
*,
1161 struct msghdr
*, size_t, int);
1162 extern int sock_no_mmap(struct file
*file
,
1163 struct socket
*sock
,
1164 struct vm_area_struct
*vma
);
1165 extern ssize_t
sock_no_sendpage(struct socket
*sock
,
1167 int offset
, size_t size
,
1171 * Functions to fill in entries in struct proto_ops when a protocol
1172 * uses the inet style.
1174 extern int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1175 char __user
*optval
, int __user
*optlen
);
1176 extern int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1177 struct msghdr
*msg
, size_t size
, int flags
);
1178 extern int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1179 char __user
*optval
, unsigned int optlen
);
1180 extern int compat_sock_common_getsockopt(struct socket
*sock
, int level
,
1181 int optname
, char __user
*optval
, int __user
*optlen
);
1182 extern int compat_sock_common_setsockopt(struct socket
*sock
, int level
,
1183 int optname
, char __user
*optval
, unsigned int optlen
);
1185 extern void sk_common_release(struct sock
*sk
);
1188 * Default socket callbacks and setup code
1191 /* Initialise core socket variables */
1192 extern void sock_init_data(struct socket
*sock
, struct sock
*sk
);
1194 extern void sk_filter_release_rcu(struct rcu_head
*rcu
);
1197 * sk_filter_release - release a socket filter
1198 * @fp: filter to remove
1200 * Remove a filter from a socket and release its resources.
1203 static inline void sk_filter_release(struct sk_filter
*fp
)
1205 if (atomic_dec_and_test(&fp
->refcnt
))
1206 call_rcu(&fp
->rcu
, sk_filter_release_rcu
);
1209 static inline void sk_filter_uncharge(struct sock
*sk
, struct sk_filter
*fp
)
1211 unsigned int size
= sk_filter_len(fp
);
1213 atomic_sub(size
, &sk
->sk_omem_alloc
);
1214 sk_filter_release(fp
);
1217 static inline void sk_filter_charge(struct sock
*sk
, struct sk_filter
*fp
)
1219 atomic_inc(&fp
->refcnt
);
1220 atomic_add(sk_filter_len(fp
), &sk
->sk_omem_alloc
);
1224 * Socket reference counting postulates.
1226 * * Each user of socket SHOULD hold a reference count.
1227 * * Each access point to socket (an hash table bucket, reference from a list,
1228 * running timer, skb in flight MUST hold a reference count.
1229 * * When reference count hits 0, it means it will never increase back.
1230 * * When reference count hits 0, it means that no references from
1231 * outside exist to this socket and current process on current CPU
1232 * is last user and may/should destroy this socket.
1233 * * sk_free is called from any context: process, BH, IRQ. When
1234 * it is called, socket has no references from outside -> sk_free
1235 * may release descendant resources allocated by the socket, but
1236 * to the time when it is called, socket is NOT referenced by any
1237 * hash tables, lists etc.
1238 * * Packets, delivered from outside (from network or from another process)
1239 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1240 * when they sit in queue. Otherwise, packets will leak to hole, when
1241 * socket is looked up by one cpu and unhasing is made by another CPU.
1242 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1243 * (leak to backlog). Packet socket does all the processing inside
1244 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1245 * use separate SMP lock, so that they are prone too.
1248 /* Ungrab socket and destroy it, if it was the last reference. */
1249 static inline void sock_put(struct sock
*sk
)
1251 if (atomic_dec_and_test(&sk
->sk_refcnt
))
1255 extern int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
1258 static inline void sk_tx_queue_set(struct sock
*sk
, int tx_queue
)
1260 sk
->sk_tx_queue_mapping
= tx_queue
;
1263 static inline void sk_tx_queue_clear(struct sock
*sk
)
1265 sk
->sk_tx_queue_mapping
= -1;
1268 static inline int sk_tx_queue_get(const struct sock
*sk
)
1270 return sk
? sk
->sk_tx_queue_mapping
: -1;
1273 static inline void sk_set_socket(struct sock
*sk
, struct socket
*sock
)
1275 sk_tx_queue_clear(sk
);
1276 sk
->sk_socket
= sock
;
1279 static inline wait_queue_head_t
*sk_sleep(struct sock
*sk
)
1281 BUILD_BUG_ON(offsetof(struct socket_wq
, wait
) != 0);
1282 return &rcu_dereference_raw(sk
->sk_wq
)->wait
;
1284 /* Detach socket from process context.
1285 * Announce socket dead, detach it from wait queue and inode.
1286 * Note that parent inode held reference count on this struct sock,
1287 * we do not release it in this function, because protocol
1288 * probably wants some additional cleanups or even continuing
1289 * to work with this socket (TCP).
1291 static inline void sock_orphan(struct sock
*sk
)
1293 write_lock_bh(&sk
->sk_callback_lock
);
1294 sock_set_flag(sk
, SOCK_DEAD
);
1295 sk_set_socket(sk
, NULL
);
1297 write_unlock_bh(&sk
->sk_callback_lock
);
1300 static inline void sock_graft(struct sock
*sk
, struct socket
*parent
)
1302 write_lock_bh(&sk
->sk_callback_lock
);
1303 sk
->sk_wq
= parent
->wq
;
1305 sk_set_socket(sk
, parent
);
1306 security_sock_graft(sk
, parent
);
1307 write_unlock_bh(&sk
->sk_callback_lock
);
1310 extern int sock_i_uid(struct sock
*sk
);
1311 extern unsigned long sock_i_ino(struct sock
*sk
);
1313 static inline struct dst_entry
*
1314 __sk_dst_get(struct sock
*sk
)
1316 return rcu_dereference_check(sk
->sk_dst_cache
, sock_owned_by_user(sk
) ||
1317 lockdep_is_held(&sk
->sk_lock
.slock
));
1320 static inline struct dst_entry
*
1321 sk_dst_get(struct sock
*sk
)
1323 struct dst_entry
*dst
;
1326 dst
= rcu_dereference(sk
->sk_dst_cache
);
1333 extern void sk_reset_txq(struct sock
*sk
);
1335 static inline void dst_negative_advice(struct sock
*sk
)
1337 struct dst_entry
*ndst
, *dst
= __sk_dst_get(sk
);
1339 if (dst
&& dst
->ops
->negative_advice
) {
1340 ndst
= dst
->ops
->negative_advice(dst
);
1343 rcu_assign_pointer(sk
->sk_dst_cache
, ndst
);
1350 __sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1352 struct dst_entry
*old_dst
;
1354 sk_tx_queue_clear(sk
);
1356 * This can be called while sk is owned by the caller only,
1357 * with no state that can be checked in a rcu_dereference_check() cond
1359 old_dst
= rcu_dereference_raw(sk
->sk_dst_cache
);
1360 rcu_assign_pointer(sk
->sk_dst_cache
, dst
);
1361 dst_release(old_dst
);
1365 sk_dst_set(struct sock
*sk
, struct dst_entry
*dst
)
1367 spin_lock(&sk
->sk_dst_lock
);
1368 __sk_dst_set(sk
, dst
);
1369 spin_unlock(&sk
->sk_dst_lock
);
1373 __sk_dst_reset(struct sock
*sk
)
1375 __sk_dst_set(sk
, NULL
);
1379 sk_dst_reset(struct sock
*sk
)
1381 spin_lock(&sk
->sk_dst_lock
);
1383 spin_unlock(&sk
->sk_dst_lock
);
1386 extern struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
);
1388 extern struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
);
1390 static inline int sk_can_gso(const struct sock
*sk
)
1392 return net_gso_ok(sk
->sk_route_caps
, sk
->sk_gso_type
);
1395 extern void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
);
1397 static inline void sk_nocaps_add(struct sock
*sk
, netdev_features_t flags
)
1399 sk
->sk_route_nocaps
|= flags
;
1400 sk
->sk_route_caps
&= ~flags
;
1403 static inline int skb_do_copy_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1404 char __user
*from
, char *to
,
1405 int copy
, int offset
)
1407 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1409 __wsum csum
= csum_and_copy_from_user(from
, to
, copy
, 0, &err
);
1412 skb
->csum
= csum_block_add(skb
->csum
, csum
, offset
);
1413 } else if (sk
->sk_route_caps
& NETIF_F_NOCACHE_COPY
) {
1414 if (!access_ok(VERIFY_READ
, from
, copy
) ||
1415 __copy_from_user_nocache(to
, from
, copy
))
1417 } else if (copy_from_user(to
, from
, copy
))
1423 static inline int skb_add_data_nocache(struct sock
*sk
, struct sk_buff
*skb
,
1424 char __user
*from
, int copy
)
1426 int err
, offset
= skb
->len
;
1428 err
= skb_do_copy_data_nocache(sk
, skb
, from
, skb_put(skb
, copy
),
1431 __skb_trim(skb
, offset
);
1436 static inline int skb_copy_to_page_nocache(struct sock
*sk
, char __user
*from
,
1437 struct sk_buff
*skb
,
1443 err
= skb_do_copy_data_nocache(sk
, skb
, from
, page_address(page
) + off
,
1449 skb
->data_len
+= copy
;
1450 skb
->truesize
+= copy
;
1451 sk
->sk_wmem_queued
+= copy
;
1452 sk_mem_charge(sk
, copy
);
1456 static inline int skb_copy_to_page(struct sock
*sk
, char __user
*from
,
1457 struct sk_buff
*skb
, struct page
*page
,
1460 if (skb
->ip_summed
== CHECKSUM_NONE
) {
1462 __wsum csum
= csum_and_copy_from_user(from
,
1463 page_address(page
) + off
,
1467 skb
->csum
= csum_block_add(skb
->csum
, csum
, skb
->len
);
1468 } else if (copy_from_user(page_address(page
) + off
, from
, copy
))
1472 skb
->data_len
+= copy
;
1473 skb
->truesize
+= copy
;
1474 sk
->sk_wmem_queued
+= copy
;
1475 sk_mem_charge(sk
, copy
);
1480 * sk_wmem_alloc_get - returns write allocations
1483 * Returns sk_wmem_alloc minus initial offset of one
1485 static inline int sk_wmem_alloc_get(const struct sock
*sk
)
1487 return atomic_read(&sk
->sk_wmem_alloc
) - 1;
1491 * sk_rmem_alloc_get - returns read allocations
1494 * Returns sk_rmem_alloc
1496 static inline int sk_rmem_alloc_get(const struct sock
*sk
)
1498 return atomic_read(&sk
->sk_rmem_alloc
);
1502 * sk_has_allocations - check if allocations are outstanding
1505 * Returns true if socket has write or read allocations
1507 static inline int sk_has_allocations(const struct sock
*sk
)
1509 return sk_wmem_alloc_get(sk
) || sk_rmem_alloc_get(sk
);
1513 * wq_has_sleeper - check if there are any waiting processes
1514 * @wq: struct socket_wq
1516 * Returns true if socket_wq has waiting processes
1518 * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
1519 * barrier call. They were added due to the race found within the tcp code.
1521 * Consider following tcp code paths:
1525 * sys_select receive packet
1527 * __add_wait_queue update tp->rcv_nxt
1529 * tp->rcv_nxt check sock_def_readable
1531 * schedule rcu_read_lock();
1532 * wq = rcu_dereference(sk->sk_wq);
1533 * if (wq && waitqueue_active(&wq->wait))
1534 * wake_up_interruptible(&wq->wait)
1538 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1539 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
1540 * could then endup calling schedule and sleep forever if there are no more
1541 * data on the socket.
1544 static inline bool wq_has_sleeper(struct socket_wq
*wq
)
1548 * We need to be sure we are in sync with the
1549 * add_wait_queue modifications to the wait queue.
1551 * This memory barrier is paired in the sock_poll_wait.
1554 return wq
&& waitqueue_active(&wq
->wait
);
1558 * sock_poll_wait - place memory barrier behind the poll_wait call.
1560 * @wait_address: socket wait queue
1563 * See the comments in the wq_has_sleeper function.
1565 static inline void sock_poll_wait(struct file
*filp
,
1566 wait_queue_head_t
*wait_address
, poll_table
*p
)
1568 if (p
&& wait_address
) {
1569 poll_wait(filp
, wait_address
, p
);
1571 * We need to be sure we are in sync with the
1572 * socket flags modification.
1574 * This memory barrier is paired in the wq_has_sleeper.
1581 * Queue a received datagram if it will fit. Stream and sequenced
1582 * protocols can't normally use this as they need to fit buffers in
1583 * and play with them.
1585 * Inlined as it's very short and called for pretty much every
1586 * packet ever received.
1589 static inline void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1593 skb
->destructor
= sock_wfree
;
1595 * We used to take a refcount on sk, but following operation
1596 * is enough to guarantee sk_free() wont free this sock until
1597 * all in-flight packets are completed
1599 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1602 static inline void skb_set_owner_r(struct sk_buff
*skb
, struct sock
*sk
)
1606 skb
->destructor
= sock_rfree
;
1607 atomic_add(skb
->truesize
, &sk
->sk_rmem_alloc
);
1608 sk_mem_charge(sk
, skb
->truesize
);
1611 extern void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1612 unsigned long expires
);
1614 extern void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
);
1616 extern int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
);
1618 extern int sock_queue_err_skb(struct sock
*sk
, struct sk_buff
*skb
);
1621 * Recover an error report and clear atomically
1624 static inline int sock_error(struct sock
*sk
)
1627 if (likely(!sk
->sk_err
))
1629 err
= xchg(&sk
->sk_err
, 0);
1633 static inline unsigned long sock_wspace(struct sock
*sk
)
1637 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
)) {
1638 amt
= sk
->sk_sndbuf
- atomic_read(&sk
->sk_wmem_alloc
);
1645 static inline void sk_wake_async(struct sock
*sk
, int how
, int band
)
1647 if (sock_flag(sk
, SOCK_FASYNC
))
1648 sock_wake_async(sk
->sk_socket
, how
, band
);
1651 #define SOCK_MIN_SNDBUF 2048
1653 * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
1654 * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
1656 #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
1658 static inline void sk_stream_moderate_sndbuf(struct sock
*sk
)
1660 if (!(sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)) {
1661 sk
->sk_sndbuf
= min(sk
->sk_sndbuf
, sk
->sk_wmem_queued
>> 1);
1662 sk
->sk_sndbuf
= max(sk
->sk_sndbuf
, SOCK_MIN_SNDBUF
);
1666 struct sk_buff
*sk_stream_alloc_skb(struct sock
*sk
, int size
, gfp_t gfp
);
1668 static inline struct page
*sk_stream_alloc_page(struct sock
*sk
)
1670 struct page
*page
= NULL
;
1672 page
= alloc_pages(sk
->sk_allocation
, 0);
1674 sk
->sk_prot
->enter_memory_pressure(sk
);
1675 sk_stream_moderate_sndbuf(sk
);
1681 * Default write policy as shown to user space via poll/select/SIGIO
1683 static inline int sock_writeable(const struct sock
*sk
)
1685 return atomic_read(&sk
->sk_wmem_alloc
) < (sk
->sk_sndbuf
>> 1);
1688 static inline gfp_t
gfp_any(void)
1690 return in_softirq() ? GFP_ATOMIC
: GFP_KERNEL
;
1693 static inline long sock_rcvtimeo(const struct sock
*sk
, int noblock
)
1695 return noblock
? 0 : sk
->sk_rcvtimeo
;
1698 static inline long sock_sndtimeo(const struct sock
*sk
, int noblock
)
1700 return noblock
? 0 : sk
->sk_sndtimeo
;
1703 static inline int sock_rcvlowat(const struct sock
*sk
, int waitall
, int len
)
1705 return (waitall
? len
: min_t(int, sk
->sk_rcvlowat
, len
)) ? : 1;
1708 /* Alas, with timeout socket operations are not restartable.
1709 * Compare this to poll().
1711 static inline int sock_intr_errno(long timeo
)
1713 return timeo
== MAX_SCHEDULE_TIMEOUT
? -ERESTARTSYS
: -EINTR
;
1716 extern void __sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
,
1717 struct sk_buff
*skb
);
1718 extern void __sock_recv_wifi_status(struct msghdr
*msg
, struct sock
*sk
,
1719 struct sk_buff
*skb
);
1721 static __inline__
void
1722 sock_recv_timestamp(struct msghdr
*msg
, struct sock
*sk
, struct sk_buff
*skb
)
1724 ktime_t kt
= skb
->tstamp
;
1725 struct skb_shared_hwtstamps
*hwtstamps
= skb_hwtstamps(skb
);
1728 * generate control messages if
1729 * - receive time stamping in software requested (SOCK_RCVTSTAMP
1730 * or SOCK_TIMESTAMPING_RX_SOFTWARE)
1731 * - software time stamp available and wanted
1732 * (SOCK_TIMESTAMPING_SOFTWARE)
1733 * - hardware time stamps available and wanted
1734 * (SOCK_TIMESTAMPING_SYS_HARDWARE or
1735 * SOCK_TIMESTAMPING_RAW_HARDWARE)
1737 if (sock_flag(sk
, SOCK_RCVTSTAMP
) ||
1738 sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
) ||
1739 (kt
.tv64
&& sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
)) ||
1740 (hwtstamps
->hwtstamp
.tv64
&&
1741 sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
)) ||
1742 (hwtstamps
->syststamp
.tv64
&&
1743 sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
)))
1744 __sock_recv_timestamp(msg
, sk
, skb
);
1748 if (sock_flag(sk
, SOCK_WIFI_STATUS
) && skb
->wifi_acked_valid
)
1749 __sock_recv_wifi_status(msg
, sk
, skb
);
1752 extern void __sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
1753 struct sk_buff
*skb
);
1755 static inline void sock_recv_ts_and_drops(struct msghdr
*msg
, struct sock
*sk
,
1756 struct sk_buff
*skb
)
1758 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
1759 (1UL << SOCK_RCVTSTAMP) | \
1760 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
1761 (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
1762 (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
1763 (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
1765 if (sk
->sk_flags
& FLAGS_TS_OR_DROPS
)
1766 __sock_recv_ts_and_drops(msg
, sk
, skb
);
1768 sk
->sk_stamp
= skb
->tstamp
;
1772 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
1773 * @sk: socket sending this packet
1774 * @tx_flags: filled with instructions for time stamping
1776 * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if
1777 * parameters are invalid.
1779 extern int sock_tx_timestamp(struct sock
*sk
, __u8
*tx_flags
);
1782 * sk_eat_skb - Release a skb if it is no longer needed
1783 * @sk: socket to eat this skb from
1784 * @skb: socket buffer to eat
1785 * @copied_early: flag indicating whether DMA operations copied this data early
1787 * This routine must be called with interrupts disabled or with the socket
1788 * locked so that the sk_buff queue operation is ok.
1790 #ifdef CONFIG_NET_DMA
1791 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1793 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1797 __skb_queue_tail(&sk
->sk_async_wait_queue
, skb
);
1800 static inline void sk_eat_skb(struct sock
*sk
, struct sk_buff
*skb
, int copied_early
)
1802 __skb_unlink(skb
, &sk
->sk_receive_queue
);
1808 struct net
*sock_net(const struct sock
*sk
)
1810 return read_pnet(&sk
->sk_net
);
1814 void sock_net_set(struct sock
*sk
, struct net
*net
)
1816 write_pnet(&sk
->sk_net
, net
);
1820 * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
1821 * They should not hold a reference to a namespace in order to allow
1823 * Sockets after sk_change_net should be released using sk_release_kernel
1825 static inline void sk_change_net(struct sock
*sk
, struct net
*net
)
1827 put_net(sock_net(sk
));
1828 sock_net_set(sk
, hold_net(net
));
1831 static inline struct sock
*skb_steal_sock(struct sk_buff
*skb
)
1833 if (unlikely(skb
->sk
)) {
1834 struct sock
*sk
= skb
->sk
;
1836 skb
->destructor
= NULL
;
1843 extern void sock_enable_timestamp(struct sock
*sk
, int flag
);
1844 extern int sock_get_timestamp(struct sock
*, struct timeval __user
*);
1845 extern int sock_get_timestampns(struct sock
*, struct timespec __user
*);
1848 * Enable debug/info messages
1850 extern int net_msg_warn
;
1851 #define NETDEBUG(fmt, args...) \
1852 do { if (net_msg_warn) printk(fmt,##args); } while (0)
1854 #define LIMIT_NETDEBUG(fmt, args...) \
1855 do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
1857 extern __u32 sysctl_wmem_max
;
1858 extern __u32 sysctl_rmem_max
;
1860 extern void sk_init(void);
1862 extern int sysctl_optmem_max
;
1864 extern __u32 sysctl_wmem_default
;
1865 extern __u32 sysctl_rmem_default
;
1867 #endif /* _SOCK_H */